An orthogonal array is formed by performing electromagnetic field analysis only once and determining a range by using the mount position and type of a capacitor and the number of capacitors as parameters to perform circuit analysis a small number of times. An estimation equation is formed by using as an index a result of the absolute value of the calculated power source impedance, and a capacitor is disposed to reduce noises by using the estimation equation.
|
1. A printed circuit board noise calculating method of calculating a power source noise of a printed circuit board mounting a semiconductor device and a decoupling capacitor by using CAD, comprising steps of:
dividing the printed circuit board into a mesh shape;
calculating an inductance matrix of all mesh positions on the printed circuit board through electromagnetic field analysis;
calculating an inductance existing between an arrangement position of the semiconductor device and an arrangement allowable position of the decoupling capacitor by using the inductance matrix;
forming an orthogonal array by using as parameters the inductance, a type of a capacitor component to be adopted as the decoupling capacitor, and the number of decoupling capacitors on the printed circuit board;
calculating parameter sets including the inductance, type of the decoupling capacitor and the number of decoupling capacitors on the printed circuit board, from the formed orthogonal array;
calculating a power source noise evaluation equation by using a power source noise of each parameter set as an index, through circuit analysis of the printed circuit board corresponding to each parameter set; and
designating a change in at least one of the arrangement position of the decoupling capacitor, the type and the number of decoupling capacitors,
wherein a power source noise corresponding to rearrangement of the decoupling capacitor caused by the change is calculated by using the calculated power source noise evaluation equation.
2. The printed circuit board noise calculating method according to
3. The printed circuit board noise calculating method according to
4. The printed circuit board noise calculating method according to
a plan view of the printed circuit board, an arrangement position of the semiconductor device, and an arrangement allowed area of the decoupling capacitor are displayed on a screen;
a change in the arrangement of the decoupling capacitor is designated by an operation of moving a pointer on the screen; and
a calculation result of the power source noise corresponding to rearrangement of the decoupling capacitor designated by the operation is displayed on the screen.
5. The printed circuit board noise calculating method according to
the electromagnetic field analysis sets ports to a plurality of semiconductor mount positions and a plurality of coupling capacitor mount positions to calculate a Y parameter matrix of the printed circuit board without mounted components, in place of the inductance;
defining capacitor Y parameter matrices corresponding to the number of mount conditions and corresponding in number to the number of mount conditions, by using the parameter sets of the orthogonal array;
adding the Y parameter matrix of the printed circuit board without mounted components obtained by the electromagnetic analysis and defined capacitor Y parameter matrices corresponding in number to the number of mount conditions to calculate impedance values under various capacitor mount conditions; and
evaluating an electric characteristic even for an unknown combination of the parameters by using the calculated evaluation equation of a polynomial equation.
|
The present application is a Divisional of U.S. application Ser. No. 12/361,761 filed Jan. 29, 2009 now U.S. Pat. No. 7,957,150, which claims priority from Japanese Patent Application No. JP 2008-039504 filed on Feb. 21, 2008, the content of which is hereby incorporated by reference into this application.
The present invention relates to a printed circuit board noise calculating method and a design support apparatus using a CAD program for a printed circuit board having semiconductor devices and decoupling capacitors constituting an information processing apparatus such as a personal computer, a server, a router and a storage.
Important guide lines for designing a printed circuit board include realizing functions provided by mounted components, reducing power source noises, reducing electromagnetic radiation from an apparatus, and the like. Reducing power source noises are particularly important from the viewpoint of ensuring stable operations of components mounted on a printed circuit board, such as semiconductor devices, analog devices and RF components. It is also important to suppress power source noises because the power source noises may become an exciting source of electromagnetic radiation noises.
A flow chart illustrated in
Component arrangement optimizing techniques are disclosed, for example, in JP-A-2001-147952 and JP-A-2004-362074. According to these techniques, after layout of a printed circuit board, models of a power source system and a ground system are formed through electromagnetic field analysis, and the two-dimensional distribution of voltage and current in the board is displayed to identify excessive noise regions. A capacitor is disposed near the excessive noise region to reduce noises.
JP-A-2000-293560 discloses techniques of speeding up electromagnetic field analysis and circuit analysis. Electromagnetic field analysis (FDTD) and circuit analysis (SPICES) are performed alternately at each step to improve an analysis precision of devices including an IC circuit.
A circuit design method disclosed in JP-A-2004-145410 allows even a person not skilled in an analog circuit to perform an optimum circuit constant design taking a circuit variation into consideration, by using Taguchi methods.
JP-A-10-207926 discloses a design approach using an orthogonal array.
The above-cited techniques are associated with the following problems.
A design flow of the techniques disclosed in JP-A-2001-147952, JP-A-2004-362074 and JP-A-2000-293560 is similar to that illustrated in
According to the technique disclosed in JP-A-2004-145410, even a person not skilled in an analog circuit can perform an optimum circuit constant design considering a circuit variation. However, there is no description on application to a printed circuit board design method aiming as low noises. It is not definite in selecting what design parameter for application to the printed circuit board design, in selecting what worst conditions, and in modeling through electromagnetic field analysis. The disclosed technique itself can not be applied to the printed circuit board design. The same issue is also associated with JP-A-10-207926.
It is an object of the present invention to provide a method of designing a printed circuit board with low power source noises, and allow even a person not skilled in the art to design at a short TAT.
The following method of the present invention settles the above-described issues. Namely, electromagnetic analysis taking a long time in a design process is performed only once. Similarly, circuit analysis taking a long time is not performed at each component rearrangement, but an orthogonal array is formed by using parameters of the type of component, a component position and the number of components. All combinations of these parameters (all parameter sets) are concentrated by forming the orthogonal array, and the circuit analysis is performed only for the concentrated parameter sets. Because of this concentration, even if the number of components increases, the number of processes will not increase exponentially. Power source noises are calculated by using as the evaluation index, the absolute values of the power source impedances of obtained by the circuit analysis at a resonance frequency and an antiresonance frequency. The evaluation index for each parameter set is formulated. With formulation, the power source impedances and power source noises of all parameter sets can be analyzed on the equation basis so that analysis time can be shortened.
Even if a designer is not a skilled designer, the designer can try an analysis with short TAT a plurality of times so that it is possible to design a printed circuit board with low noises.
By applying the present invention to design of a multilayer board of an information processing apparatus such as a personal computer, a server, a router and a storage, and home information appliances, a design cost of the apparatus can be lowered. Since a try can be performed a number of times, an optimum arrangement realizing lower power source noises can be derived, and a stable operation and low radiation noise characteristic of a product can thus be acquired.
Other objects, features and advantages of the invention will become apparent from the following description of the embodiments of the invention taken in conjunction with the accompanying drawings.
[First Embodiment]
The first embodiment will be described with reference to
In a substrate information inputting process 110, design information of a printed circuit board necessary for printed circuit board design is input, including outer shape information of the printed circuit board, the number of layers, shape information of via holes and the like. At this stage, a wiring design rule and the like not shown are prepared in advance, including circuit connection information (net list), component list information, a wiring inhibition shortest distance, and the like. By using the net list and component list, a component arranging process 120 is executed. The component list includes information on each semiconductor device providing a function and information on each decoupling component to be connected to each semiconductor device to stabilize the power source and reduce power source noises. In the component arranging process 120, components having a large size such as semiconductor devices, components having a large consumption power and components being sensitive and easy to improperly respond to power source noises are arranged tentatively, and decoupling capacitors are arranged tentatively around these components. In this process, a board layout is performed in some cases, including setting an outer shape of the printed circuit board, dividing power source wirings of different power sources.
Next, an electromagnetic analyzing process 130 is performed relative to the shapes of a power source layer and a ground layer, or a plane (hereinafter simply called a power source plane), disposed in the printed circuit board. The power source plane is meshed and an inductance matrix (Lk) of each mesh is obtained.
Next, a parameter set calculating process 140 forms an orthogonal array (p) by using as parameters the type (t) of a decoupling capacitor, the number (n) of decoupling capacitors and an inductance corresponding to arrangement. The number of parameter sets formed by the orthogonal array is N. This number N is characterized in that N is much smaller than all combinations of the parameters.
A circuit analyzing process 150 forms an equivalent circuit model corresponding to the parameter sets to perform electric circuit analysis of the power source plane and calculate the power source impedances (Z11) of the whole printed circuit board as seen from a power source terminal of each semiconductor device. The power source impedances (Z11) include the impedances at a resonance frequency and an antiresonance frequency. There is a property that the impedance Z11 at the antiresonance frequency becomes high. Therefore, the absolute value |Z11| of the power source impedance at the antiresonance frequency (f0) is calculated, and is used as the electric characteristic evaluation index corresponding to a parameter set. These calculating processes 150 and 160 are repeated the number (N) of parameter sets to obtain evaluation indices |Z11| of the parameter sets.
Next, an electric characteristic evaluation index formulating process 180 formulates a calculation equation for the power source impedance applicable to all combinations of the parameters, by using the evaluation index |Z11|. Generally, this calculation equation can be calculated at high speed because it is represented by a polynomial equation having an order of (level number—1) and using the parameters as variables.
Next, a noise voltage estimating process 190 calculates power source noises relative to component arrangement. This calculation is obtained from a product of the absolute value |Z11| of the power source impedance for each parameter set calculated previously and power source noise current components (I(f)) generated by the semiconductor device at an antiresonance frequency (f). Namely, a noise voltage is given by:
Noise voltage: Vn(f)=|Z11|×I(f) (1)
A judging process 200 judges whether the noise voltage (Vn) is not larger than a target value set by a user. If the noise voltage is larger than the target value, a component rearranging process 210 rearranges the component. In accordance with new component arrangement information, the process 190 estimates the noise voltage. The processes 190 to 210 are repeated until the noise voltage becomes not larger than the target value. The process is terminated after the noise voltage becomes not larger than the target value.
Next, each process will be described in detail.
First, in order to facilitate understanding, an example of a cross section of a printed circuit board as a design target is illustrated in
As described above, the board information input process 110 is the same as a general printed circuit board design process, and the duplicated description will be omitted.
The component arranging process 120 tentatively arranges components such as semiconductor devices 11 and 12 and decoupling capacitors 21 and 22 on the bear board 51 of a printed circuit board illustrated in
In the electromagnetic field analyzing process 130, as illustrated in
Next, the parameter set calculating process 140 calculates an inductance (L) existing between the position of a semiconductor device and the position where a capacitor can be arranged, by using the inductance matrices (Lk) obtained by the electromagnetic analyzing process 130.
Description will be made on a case wherein the semiconductor device 11 illustrated in
A portion indicated at a level 2 represents an equivalent circuit of a semiconductor package, and an inductance, a resistance and a capacitor C3 are connected in a character shape of alphabet “H”. A portion indicated at a level 3 corresponds to an equivalent circuit of the printed circuit board 51, and similarly a capacitance C4 and other parasitic components are connected in a character shape of “H”. Similarly, a portion indicated at a level 4 corresponds to an equivalent circuit of a power feeding circuit, and external wirings and a power source to be connected via connectors of the power source wiring on the printed circuit board. An equivalent circuit of the semiconductor device 11 is represented by 130-1 indicated by a dotted line, an equivalent circuit of the capacitors 21 and 22 is represented by 130-2, and an equivalent circuit of a portion from the level 1 to level 3 is represented by 130-3.
An impedance ZL of an inductance L and impedance (ZC) of a capacitance C are given by the following equations:
ZL=jωL (2)
ZC=1/(jωC) (3)
and a resonance frequency (ω0) of LC is given by the following equation:
ω0=1/√(LC) (4)
wherein j is a pure imaginary number, ω is an angular frequency, and there is a relation of ω=2πf relative to the frequency f.
Therefore, inclinations of the impedances of the inductance (L) and capacitance (C) relative to the frequency (f) are opposite. In
Similarly, the inductance is different on the digit unit basis. In the case of the inductance, since the area of a current flow path is proportional to crossing magnetic fluxes, it is possible to consider that the inductance is proportional to the area of the current flow path. For example, about 1 cm at the level 1 (LSI), about several cm at the level 2 (PKG), and about several ten cm at the level 3 (PCB). As above, since the printed circuit board has capacitances and inductances different on the digit unit basis at respective levels, the power source impedances are irregular also on the digit unit basis. Impedances at each resonance frequency and each antiresonance frequency depend upon a Q value (quality factor of each loop circuit), and a parasitic resistance value of each loop circuit depends upon a conductance value.
As described above, the inductance of the absolute value |Z11| of the power source impedance changes with a distance (an area of crossed fluxes) between the semiconductor device and a decoupling capacitor. Therefore, the power source impedance and power source noises change with a position at which the decoupling capacitor is arranged. Namely, in the equivalent circuit in
Description will be made on the parameter set calculating process 140 illustrated in
Calculated at Step 141 is a range of an inductance parameter (L) relative to the arrangement of the decoupling capacitor. This calculating Step 141 is executed as illustrated in
At Step 1412, the coordinates values of the substrate are converted into mesh coordinate values of the equivalent circuit. Namely, assuming that the position of a power source pin of the semiconductor device 11 is represented by the coordinate values 1 and the position of the capacitor is represented by the coordinate values 2, by using the information on coordinate values of components on the printed circuit board, the coordinate values 1 and 2 are converted into mesh coordinate values (i1, j1) and (i2, j2) on the meshes of the equivalent circuit 130-a, respectively.
At Step 1413, a minimum inductance (Lmin) is calculated. In this calculation, an inductance (L) between two mesh coordinate values (i1, j1) and (i2, j2) corresponding to the shortest distance is calculated from the equivalent circuit 130-a illustrated in
With these Steps, the range of the inductance parameter (L) can be designated.
Similarly, a range of a parasitic inductance is calculated at Step 142 illustrated in
As illustrated in
Next, at Step 143 illustrated in
At Step 144, an orthogonal array is formed by using three parameters (L, t, n) determined at Steps 141 to 143. For example, if each parameter has a maximum value, a minimum value and a median value and the orthogonal array is formed at the third level, nine parameter sets (p) are determined which are 1/third a total number of combinations of 3^3=27. A circuit analysis time can therefore be reduced considerably. The details of an orthogonal array are given in JP-A-2004-145410.
At Step 150 illustrated in
Calculation may be performed by using an equivalent circuit 130-a of the printed circuit board 51 as illustrated in
Calculation at Step 150 in
At Step 180, an estimation equation for the absolute value |Z11| index of the power source impedance is calculated at each parameter (L, t, n) of the parameter sets (p). This calculation method is detailed in JP-A-10-207926. By using the parameters as variables, the Z11 estimation equation (5) is given by:
Z11=|Z11(p)|=|Z11(L, t, n)|=a0+Σ(ai·L^i+bi·t^i+ci·n^i) (i=1 . . . 2) (5)
having an effective precision in the range of Lmin<L<Lmax, ESLmin<t<ESLmax, and nmin<n<nmax. In this equation, a0 is a constant, ai, bi and ci are coefficients of each parameter (L, t, n), and i is an integral number, in equation (5) i is from 1 to 2. If necessary, an interaction such as a product of L and n may be defined and a secondary term may be added to the estimation equation to improve the precision. This Step 180 can obtain the absolute value |Z11| of the power source impedance as the electric characteristic evaluation index, through arithmetic calculation without electromagnetic field analysis, for every and all combinations of the parameters if in the range of each parameter.
At Step 190 in
This estimation is calculated by substituting Z11 for each parameter set (p) into the equation (1). Since this calculation is arithmetic calculation, calculation can be performed at high speed.
It is judged at Step 200 in
As illustrated in
Namely, by performing electromagnetic field analysis and circuit analysis for the final component arrangement, it becomes possible to verify at high precision whether the power source noises for the final component arrangement are actually lower than the target value. Therefore, even if the estimation equation obtained at Step 180 contains many errors, the design precision can be ensured because electromagnetic field analysis and circuit analysis are performed at Step 220.
After the detailed verification is performed at Step 220, if it is judged at Step 230 that the noise voltage is larger than the target value, the flow returns to Step 140 whereat parameter sets are calculated again by narrowing the parameter range. Concentration of parameter sets can therefore be performed with short TAT.
In this embodiment, the power source noises are estimated by using the absolute value |Z11| of the power source impedance as the electric characteristic evaluation index and the equation (1). Instead of this method, the power source noises may be calculated by the circuit analysis 150, and the maximum value (Vn) of variable noises may be used as the electric characteristic evaluation index to be obtained as Step 160. Formulation at Step 180 using Vn is given from the same concept as the equation (5) by:
Vn=Vn(p)=Vn(L, t, n)=a0 +Σ(ai·L^i+bi·t^i+ci·n^i) (6)
where i=1 . . . 2, Lmin<L<Lmax, ESLmin<t<ESLmax, and nmin<n<nmax.
In this case, at Step 190, arithmetic calculation of the equation (6) is performed. Similarly, the parameter set (p) is not limited only to (L, t, n), but a size of a package, the number of balls, a signal/ground ratio and the like may be selected. This arrangement improves the degree of design freedom, and ensures the essence of short TAT design without repetition of circuit analysis.
[Second Embodiment]
With reference to
A different point of
Z11(f)=ZC2//ZC3//ZC4//ZL3//ZL4 (7)
where a symbol “//” means a total impedance of parallel connection components. Each value of ZC2, ZC3, ZC4, ZL3 and ZL4 is formulated by a polynomial equation of each parameter set (p) like the equation (5). For example, ZC2 is expressed by the following polynomial equation:
ZC2=a0+Σ(ai·L^i+bi·t^i+ci·n^i) (8)
where i=1 . . . 2. The absolute values |Z11| at a plurality of frequencies (f) can be arithmetically calculated at high speed by using parameters of ZC2, ZC3, ZL3 or the like expressed by a polynomial equation.
The power source noise waveform Vgnd(t) can be calculated by:
Vgnd(t)=Σ(Z11(fi)×I(fi)) (9)
where i=1, . . . , k. Namely, the power source noise waveform can be calculated by calculating a product of a current value of a noise current model and the power source impedance at each selected one of k antiresonance frequencies. This calculation is also arithmetic calculation using a polynomial equation, software allowing real time display can therefore be implemented.
Since the target impedance (Ztarget) is displayed in the area 82, a user can design an arrangement of the capacitor 22 to move near to the target. Similarly, as a noise voltage target value (Vtarget) is displayed, similar effects can be obtained.
The correspondence between each Step in
The number (n) of capacitors can be increased by GUI operation. A designer sequentially adds a capacitor one by one to increase the number (n).
According to the second embodiment, even if a designer is not a skilled designer, the type, mount position and the number of decoupling capacitors can be selected by only the operation on the screen while the designer monitors the target power source impedance and voltage waveform, so that design can be completed in a short time. This can be achieved by using the estimation equation (5 or 6) without repeating circuit analysis at each component rearrangement.
[Third Embodiment]
The third embodiment will be described by using
Description will be made on the reason why the calculation is speeded up more that the circuit calculation, by using Y parameters.
First, description will be made on how a change in the power feeding system impedance when a capacitor component is replaced is calculated. With this calculation approach, in the state that mount component information is not added to a target printed circuit board as illustrated in
The Y parameters pf of a capacitor can be calculated easily from the equivalent circuit of the capacitor illustrated in
[0] represents a zero matrix, and [Yc] represents a diagonal matrix of decoupling capacitors.
In this manner, a plurality of capacitor admittance matrices Yc corresponding to the capacitor mount conditions are prepared and calculated on the basis of the parameter sets obtained at Step 140 to obtain the admittance matrix of the whole circuit corresponding to the capacitor mount conditions. Therefore, the Y parameters corresponding to various capacitor mount conditions as seen from each LSI can be obtained to calculate the Z11 from the Y parameters.
Although the Y parameters are used, other frequency parameters such as S parameters and Z parameters (impedance parameters) may also be used to perform similar matrix calculations.
It should be further understood by those skilled in the art that although the foregoing description has been made on embodiments of the invention, the invention is not limited thereto and various changes and modifications may be made without departing from the spirit of the invention and the scope of the appended claims.
Uematsu, Yutaka, Osaka, Hideki
Patent | Priority | Assignee | Title |
10785867, | Sep 25 2018 | International Business Machines Corporation | Automatic determination of power plane shape in printed circuit board |
Patent | Priority | Assignee | Title |
7114132, | Apr 20 2001 | NEC Corporation | Device, system, server, client, and method for supporting component layout design on circuit board, and program for implementing the device |
7120885, | Jan 27 2000 | Matsushita Electric Industrial Co., Ltd. | Computer aided design apparatus for aiding design of a printed wiring board to effectively reduce noise |
7168058, | Jun 16 2003 | NEC Corporation | Printed circuit wiring board designing support device, printed circuit board designing method, and its program |
7240319, | Feb 19 2003 | DIVSYS INTERNATIONAL LLC | Apparatus, system, method, and program for facilitating the design of bare circuit boards |
7353483, | Feb 20 2004 | Godo Kaisha IP Bridge 1 | Element arrangement check device and printed circuit board design device |
7428717, | Jan 30 2006 | XILINX, Inc. | Integrated system noise management—system level |
20010018761, | |||
20020176989, | |||
20060282811, | |||
20070062726, | |||
JP10207926, | |||
JP2000293560, | |||
JP2001147952, | |||
JP2004145410, | |||
JP2004362074, |
Executed on | Assignor | Assignee | Conveyance | Frame | Reel | Doc |
Jun 07 2011 | Hitachi, Ltd. | (assignment on the face of the patent) | / |
Date | Maintenance Fee Events |
Mar 12 2014 | ASPN: Payor Number Assigned. |
Jun 30 2016 | M1551: Payment of Maintenance Fee, 4th Year, Large Entity. |
Jul 02 2020 | M1552: Payment of Maintenance Fee, 8th Year, Large Entity. |
Jul 03 2024 | M1553: Payment of Maintenance Fee, 12th Year, Large Entity. |
Date | Maintenance Schedule |
Jan 15 2016 | 4 years fee payment window open |
Jul 15 2016 | 6 months grace period start (w surcharge) |
Jan 15 2017 | patent expiry (for year 4) |
Jan 15 2019 | 2 years to revive unintentionally abandoned end. (for year 4) |
Jan 15 2020 | 8 years fee payment window open |
Jul 15 2020 | 6 months grace period start (w surcharge) |
Jan 15 2021 | patent expiry (for year 8) |
Jan 15 2023 | 2 years to revive unintentionally abandoned end. (for year 8) |
Jan 15 2024 | 12 years fee payment window open |
Jul 15 2024 | 6 months grace period start (w surcharge) |
Jan 15 2025 | patent expiry (for year 12) |
Jan 15 2027 | 2 years to revive unintentionally abandoned end. (for year 12) |